Development of Pneumonia Patient Classification Model Using Fair Federated Learning

Worldwide, pneumonia has been a major problem for the past few centuries. Currently, medical staff are having a hard time due to the increase in COVID-19 in many countries. Chest X-ray is the most common method for screening and diagnosing chest diseases. However, there are difficulties in building the model due to data confidentiality between patients and hospitals and problems with collecting large amounts of data within hospitals. As a solution to this, we propose FFLFCN, which uses federated learning and deep learning to diagnose pneumonia. FFLFCN built a central model by training local models in multiple hospitals with their own data while maintaining the privacy of patient data. In this paper we aim to inspire federated learning research for pneumonia patients using FFLFCN, and achieve improved AUC and shorter learning time than before.